Contact Us

Product Sense: A Hidden Lynchpin in Data Science and AI

Alex Olden
Senior Data Scientist

Vigilantly monitoring eraser supplies didn’t make the list of critical tasks when I started teaching, but it should have. I had already learned about classroom management, lesson planning, and the unique rules and rhythms of my school. To be sure, these components are essential for successful teaching. But once I got the keys to my own classroom, I quickly found that a pack of Pentel Hi-Polymers could be my best friend. Little things matter, as we used to say. 

Scenario A: students enter my room, find their seats, and start passing up homework. Everyone is on-task and moving along nicely when Chris notices the eraser on his pencil is spent. He raises his hand, I call on him, we discuss erasers, etc. None of this is Chris’s fault, of course, but the rest of the class astutely notices that I’m distracted. It’s a great time to pass notes, attempt anonymous noises, or derail the class with one’s best material. To paraphrase Mike Tyson, everyone has a lesson plan until a student turns their pencils into drumsticks. 

Now picture Scenario B: I notice Chris off-task and a bit perplexed, see that he needs an eraser, and hand him one. Class proceeds as intended. End scene. 

Looking back, I understand that my gravitation toward Scenario B was a major part of effective teaching because it entailed knowing and anticipating the needs of customers. Disclaimer: this insight should not be taken as reductive, transactional, or impersonal. My teaching relied heavily on strong relationships that took time to build with students, and no student is won over or motivated by mere erasers. But if we angle our brains for a moment and think of students as customers and teachers as service providers, we can glimpse me honing an essential skill for my current work as a data scientist: product sense. 

Like many things in the world of data science (including the term itself), product sense means different things to different people. I think of it in two main pieces. The first one is opportunity identification: if we get data X, we can build product Y for customer segment Z. The second part is product evaluation. Here we answer essential questions about the health and value of the product itself: Which customers are using it, and how? What happens if we change the product in a certain way? Overall, product sense is rooted in seeking ways to serve customers by using available data, and ways to monitor the performance of the resulting products and services. 

I once heard a product expert say that if you’re designing a basketball and start by thinking about the ball, you’ve already gone wrong. You must start by thinking about the basketball player. AI is no different: we need to start by thinking about the end user of the tools we build. Building a recommendation system to show new products to customers is exciting and impressive, but first we need to learn about the customers who’ll be receiving those recommendations. 

Due diligence up front is not optional when it comes to building AI tools that add genuine value for businesses and their customers. Such work starts with thorough data exploration and analysis, relevant opportunity identification, and thoughtful metric design to help us monitor performance. Otherwise, we end up with the infamous solution in search of a problem – or product in search of a customer. The antidote here is product sense, always being the teacher ready with an eraser. 

By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.
X Icon